An important trend in underwater acoustics is the use of large arrays with many elements and small resolution cells in an effort to detect weak signals. Frequently, the background noise is dominated by multiple moving surface ships that act like discrete interferers. These interferers may move through many spatial resolution cells in the time required to obtain sufficient "snapshots" for covariance matrix estimation. The non-stationarity of the background limits the performance of adaptive algorithms. Reduced degree-of-freedom approaches that can track and rapidly null interferers before they move too far are of great interest.

The paper presents a new class of algorithms in which a low rank dynamically changing interference subspace is estimated and tracked using a short-term average. An orthogonal projection is then applied to the data to null components in the interference subspace. A second stage of adaptive beamforming that uses a longer averaging time is then applied to the projected data. Unlike most subspace approaches that work for strong signals, then is no requirement that the signal be contained in the interference subspace. It is shown how multiple interferers moving rapidly across beams can be tracked and nulled to reveal the presence of a weak target not visible with conventional, adaptive or usual sub-space methods. The effectiveness of multirate adaptive beamforming is illustrated by examples. It is shown that moving interferers are effectively nulled to allow detection of weak signals when other methods including MVDR, conventional and signal subspace fail.

[Back to Session 11] [Back to Technical Program]